framework,version,device,op_name,kernel_source,mla_dtype,kv_cache_dtype,num_heads,batch_size,isl,tp_size,step,latency
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1,1,1,0,0.6264399960637093
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1,4,1,0,0.631348006427288
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1,8,1,0,0.6289639919996262
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1,16,1,0,0.6369320079684258
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1,32,1,0,0.6276680007576942
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1,64,1,0,0.6341720074415207
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1,128,1,0,0.6217680051922798
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1,256,1,0,0.6268080025911331
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1,512,1,0,0.637584000825882
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1,1,1,0,0.6059560105204582
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1,1024,1,0,0.6249319911003113
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1,4,1,0,0.6237560138106346
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1,2048,1,0,0.6602599993348122
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1,8,1,0,0.6031000018119812
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1,4096,1,0,0.7645359933376312
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1,16,1,0,0.6069159954786301
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1,32,1,0,0.61166001111269
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1,8192,1,0,1.1963040083646774
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1,64,1,0,0.6080360114574432
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1,128,1,0,0.6054599955677986
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1,16384,1,0,2.7523439824581146
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1,256,1,0,0.691596008837223
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,2,4,1,0,0.6715120002627373
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1,512,1,0,0.6464200094342232
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,2,8,1,0,0.6579599976539612
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1,1024,1,0,0.6652239933609962
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1,1,1,0,0.61507198959589
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1,2048,1,0,0.7163679897785187
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,2,16,1,0,1.0886680036783218
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1,4,1,0,0.6286880001425743
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1,4096,1,0,0.9570439904928207
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,2,32,1,0,1.1096560060977936
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1,8,1,0,0.6364639848470688
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1,8192,1,0,1.818575993180275
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,2,64,1,0,1.1208279877901077
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1,16,1,0,0.6119880005717278
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1,32,1,0,0.6185799986124039
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,2,128,1,0,1.1908079981803894
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1,64,1,0,0.6181560084223747
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,2,256,1,0,1.2100559920072556
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1,128,1,0,0.604200005531311
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1,16384,1,0,4.988067924976349
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,2,512,1,0,1.2160120010375977
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1,256,1,0,0.6148639991879463
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,2,4,1,0,0.6270639970898628
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,2,1024,1,0,1.139820009469986
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1,512,1,0,0.6364319995045662
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,2,8,1,0,0.6154199913144112
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1,1024,1,0,0.6867320016026497
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,2,2048,1,0,1.137348011136055
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,2,16,1,0,0.6215800046920776
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1,2048,1,0,0.8603160008788109
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,2,32,1,0,0.622019998729229
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,2,4096,1,0,1.1736959889531136
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1,4096,1,0,1.3597799837589264
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,2,64,1,0,0.6182879954576492
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,2,8192,1,0,1.8912480026483536
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,2,128,1,0,0.6110360026359558
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,2,256,1,0,0.6281840056180954
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1,8192,1,0,3.3596599996089935
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,2,512,1,0,0.6528719961643219
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,2,1024,1,0,0.689351998269558
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,2,16384,1,0,5.362823903560638
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,2,2048,1,0,0.8424039930105209
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,4,4,1,0,0.6382480040192604
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,2,4096,1,0,1.3881959915161133
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,4,8,1,0,0.6417959928512573
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,4,16,1,0,0.6655599921941757
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,4,32,1,0,0.6499039903283119
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,2,8192,1,0,3.4105679988861084
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1,16384,1,0,10.893983960151672
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,4,64,1,0,0.6582679897546768
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,2,4,1,0,0.6214199960231781
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,4,128,1,0,0.6536440029740334
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,2,8,1,0,0.6270400062203407
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,4,256,1,0,0.6549440026283264
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,2,16,1,0,0.6159640029072762
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,4,512,1,0,0.6747959926724434
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,2,32,1,0,0.6313199996948242
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,4,1024,1,0,0.7140320017933846
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,2,64,1,0,0.6348000019788742
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,4,2048,1,0,0.883140005171299
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,2,128,1,0,0.6394079998135567
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,2,16384,1,0,10.79781198501587
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,2,256,1,0,0.6592039987444878
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,4,4096,1,0,1.4582319855690002
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,4,4,1,0,0.6301199868321419
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,2,512,1,0,0.7000879943370819
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,4,8,1,0,0.6196919903159142
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,2,1024,1,0,0.8246679976582527
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,4,16,1,0,0.6171080097556114
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,4,8192,1,0,3.5806040167808533
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,2,2048,1,0,1.1657959967851639
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,4,32,1,0,0.6249440014362335
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,4,64,1,0,0.6255439966917038
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,2,4096,1,0,2.5264119803905487
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,4,128,1,0,0.6214519962668419
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,4,256,1,0,0.647211991250515
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,4,512,1,0,0.7077679932117462
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,4,1024,1,0,0.7873559817671776
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,4,2048,1,0,1.1845159828662872
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,2,8192,1,0,6.944664001464844
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,4,16384,1,0,11.005067944526672
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,8,4,1,0,0.6430640071630478
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,4,4096,1,0,2.542871981859207
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,8,8,1,0,0.6567559912800789
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,8,16,1,0,0.6457280069589615
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,8,32,1,0,0.6536120027303696
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,8,64,1,0,0.6476559862494469
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,8,128,1,0,0.6471880078315735
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,4,8192,1,0,6.968252003192902
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,8,256,1,0,0.6779520064592361
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,8,512,1,0,0.703220009803772
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,8,1024,1,0,0.8228159993886948
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,8,2048,1,0,1.2357160151004791
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,8,4096,1,0,2.682280033826828
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,2,16384,1,0,21.108787775039673
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,4,4,1,0,0.6363120004534721
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,4,8,1,0,0.6381999999284744
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,4,16,1,0,0.6284520104527473
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,4,32,1,0,0.6222120001912117
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,8,8192,1,0,7.152112007141113
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,4,64,1,0,0.6423159912228584
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,4,128,1,0,0.6559520065784454
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,4,16384,1,0,21.344280242919922
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,4,256,1,0,0.6968639940023422
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,8,4,1,0,0.6310519948601723
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,4,512,1,0,0.7878080010414124
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,8,8,1,0,0.6394560039043427
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,4,1024,1,0,1.0415599942207336
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,8,16,1,0,0.6317159906029701
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,8,32,1,0,0.6272599995136261
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,4,2048,1,0,2.0878719985485077
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,8,64,1,0,0.6296759992837906
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,8,128,1,0,0.6420999988913536
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,8,256,1,0,0.7072400003671646
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,4,4096,1,0,4.869247913360596
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,8,512,1,0,0.7641839981079102
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,8,1024,1,0,1.150712013244629
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,8,2048,1,0,2.114467978477478
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,8,16384,1,0,22.310652017593384
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,16,4,1,0,0.67170000821352
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,16,8,1,0,0.6572440043091774
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,16,16,1,0,0.6685040071606636
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,8,4096,1,0,4.818620026111603
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,16,32,1,0,0.6643599942326546
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,4,8192,1,0,13.897220015525818
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,16,64,1,0,0.6592000052332878
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,16,128,1,0,0.6895239949226379
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,16,256,1,0,0.7219960018992424
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,16,512,1,0,0.804131992161274
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,16,1024,1,0,1.1434239894151688
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,16,2048,1,0,2.275400012731552
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,8,8192,1,0,14.05049216747284
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,16,4096,1,0,5.165607988834381
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,32,4,1,0,0.6786920055747032
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,32,8,1,0,0.6823520138859749
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,32,16,1,0,0.6929680034518242
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,32,32,1,0,0.6822720021009445
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,32,64,1,0,0.7155320048332214
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,32,128,1,0,0.7410560101270676
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,32,256,1,0,0.8304199948906898
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,32,512,1,0,1.116923987865448
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,32,1024,1,0,2.085819959640503
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,32,2048,1,0,4.395632088184357
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,4,16384,1,0,43.469143867492676
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,8,4,1,0,0.6382919996976852
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,8,8,1,0,0.647087998688221
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,8,16,1,0,0.6352239921689034
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,8,32,1,0,0.6409360095858574
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,32,4096,1,0,10.939671993255615
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,8,64,1,0,0.6609399989247322
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,8,128,1,0,0.7013439983129501
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,64,4,1,0,0.735660009086132
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,8,256,1,0,0.7900279983878136
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,8,16384,1,0,43.35722017288208
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,64,8,1,0,0.7380640059709549
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,8,512,1,0,0.998816005885601
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,64,16,1,0,0.7570199966430664
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,16,4,1,0,0.6632039919495583
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,64,32,1,0,0.7706280127167702
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,16,8,1,0,0.6625319942831993
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,8,1024,1,0,1.8775199949741364
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,64,64,1,0,0.8024679943919182
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,16,16,1,0,0.6364440023899078
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,64,128,1,0,0.8791320025920868
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,16,32,1,0,0.6476439982652664
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,16,64,1,0,0.6755999997258186
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,64,256,1,0,1.1372160166502
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,8,2048,1,0,3.8761320412158966
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,16,128,1,0,0.7063919976353645
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,16,256,1,0,0.7806959897279739
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,64,512,1,0,2.0266120433807373
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,16,512,1,0,1.0325319916009903
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,16,1024,1,0,1.9174080044031143
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,64,1024,1,0,3.9870240092277527
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,8,4096,1,0,9.937680006027222
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,16,2048,1,0,4.084184050559998
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,64,2048,1,0,8.958667874336243
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,128,4,1,0,0.8411200046539307
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,128,8,1,0,0.861672006547451
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,16,4096,1,0,10.158543825149536
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,128,16,1,0,0.8865840062499046
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,32,4,1,0,0.674015998840332
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,128,32,1,0,0.9126359820365906
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,32,8,1,0,0.6730840057134628
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,128,64,1,0,1.0398159995675087
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,32,16,1,0,0.6828079968690872
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,32,32,1,0,0.6983280032873154
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,128,128,1,0,1.2427280098199844
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,32,64,1,0,0.745168000459671
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,32,128,1,0,0.7955999970436096
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,128,256,1,0,2.0877000093460083
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,32,256,1,0,1.0716760158538818
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,32,512,1,0,1.8386759907007217
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,8,8192,1,0,27.872559547424316
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,128,512,1,0,3.8625520169734955
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,32,1024,1,0,3.586972028017044
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,128,1024,1,0,8.122992038726807
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,256,4,1,0,1.1059679985046387
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,32,2048,1,0,8.30978000164032
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,256,8,1,0,1.1296399980783463
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,256,16,1,0,1.1732320189476013
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,256,32,1,0,1.2585639953613281
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,256,64,1,0,1.4839880019426346
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,256,128,1,0,2.26800400018692
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,256,256,1,0,3.940580040216446
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,32,4096,1,0,20.41127586364746
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,64,4,1,0,0.720928005874157
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,64,8,1,0,0.7418480068445206
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,64,16,1,0,0.756432019174099
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,256,512,1,0,7.759424030780792
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,64,32,1,0,0.8054879978299141
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,64,64,1,0,0.8701839968562126
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,512,4,1,0,1.6634400188922882
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,64,128,1,0,1.0696480125188828
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,512,8,1,0,1.6741360127925873
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,64,256,1,0,1.8282959908246994
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,512,16,1,0,1.7790759950876236
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,64,512,1,0,3.4214640259742737
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,512,32,1,0,2.000563994050026
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,512,64,1,0,2.741627961397171
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,64,1024,1,0,7.294428110122681
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,512,128,1,0,4.336199998855591
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,512,256,1,0,7.790615975856781
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,64,2048,1,0,16.634495973587036
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1024,4,1,0,2.802828013896942
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,128,4,1,0,0.8979079946875572
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,128,8,1,0,0.9072319939732552
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1024,8,1,0,2.9060560166835785
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,128,16,1,0,0.9782639816403389
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,8,16384,1,0,88.37145519256592
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,128,32,1,0,1.0377279967069626
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,16,4,1,0,0.6531759947538376
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1024,16,1,0,3.094484031200409
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,128,64,1,0,1.2333840131759644
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,16,8,1,0,0.6428399980068207
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,16,16,1,0,0.6488479971885681
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,128,128,1,0,1.918511986732483
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,16,32,1,0,0.6770200058817863
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1024,32,1,0,3.8782760202884674
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,16,64,1,0,0.7214559987187386
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,16,128,1,0,0.8038760051131248
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,128,256,1,0,3.4081080555915833
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,16,256,1,0,0.9915519952774048
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,16,512,1,0,1.7842279970645905
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1024,64,1,0,5.375551998615265
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,128,512,1,0,6.913584053516388
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,16,1024,1,0,3.5181999802589417
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,16,1024,128,1,0,8.588032007217407
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,16,2048,1,0,8.099271893501282
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,128,1024,1,0,14.898604035377502
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,256,4,1,0,1.2415200024843216
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,256,8,1,0,1.2923599928617477
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,256,16,1,0,1.3435599654912949
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,256,32,1,0,1.544388011097908
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,16,4096,1,0,19.848227977752686
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,256,64,1,0,2.226763993501663
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,32,4,1,0,0.6602479889988899
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,32,8,1,0,0.6821840032935143
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,256,128,1,0,3.5843720138072968
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,32,16,1,0,0.6943719983100891
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,32,32,1,0,0.8871159926056862
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,32,64,1,0,0.8387600034475327
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,32,128,1,0,1.0048799887299538
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,256,256,1,0,6.81521201133728
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,32,256,1,0,1.7477879971265793
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,32,512,1,0,3.291671961545944
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,32,1024,1,0,7.306547939777374
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,256,512,1,0,13.9688960313797
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,512,4,1,0,1.9257400184869766
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,512,8,1,0,1.9947400093078613
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,512,16,1,0,2.1683399975299835
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,32,2048,1,0,16.326567888259888
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,512,32,1,0,2.918924003839493
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,512,64,1,0,4.144112050533295
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,512,128,1,0,7.008935987949371
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,512,256,1,0,13.661495923995972
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1024,4,1,0,3.302743971347809
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,32,4096,1,0,42.12363243103027
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1024,8,1,0,3.4929400086402893
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,64,4,1,0,0.8309160023927689
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,64,8,1,0,0.8258999884128571
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,64,16,1,0,0.8743400052189827
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1024,16,1,0,4.137899994850159
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,64,32,1,0,0.9566560089588165
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,64,64,1,0,1.1019359976053238
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,64,128,1,0,1.787203997373581
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1024,32,1,0,5.457548022270203
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,64,256,1,0,3.1997920274734497
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1024,64,1,0,8.224796056747437
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,64,512,1,0,6.403999924659729
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,64,1024,1,0,14.280464172363281
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,32,1024,128,1,0,14.175164103507996
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,64,2048,1,0,32.432151794433594
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,128,4,1,0,1.0384519845247269
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,128,8,1,0,1.0808839797973633
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,128,16,1,0,1.1767280101776123
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,128,32,1,0,1.334319993853569
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,128,64,1,0,1.9996599853038788
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,128,128,1,0,3.2843320071697235
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,128,256,1,0,6.200919985771179
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,128,512,1,0,13.43210780620575
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,128,1024,1,0,28.539127826690674
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,256,4,1,0,1.5473519712686539
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,256,8,1,0,1.6186400204896927
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,256,16,1,0,1.7836959958076477
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,256,32,1,0,2.4233680069446564
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,256,64,1,0,3.6751959919929504
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,256,128,1,0,6.3430920243263245
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,256,256,1,0,12.850883960723877
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,256,512,1,0,26.596303939819336
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,512,4,1,0,2.5424280166625977
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,512,8,1,0,2.6813879907131195
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,512,16,1,0,3.275743991136551
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,512,32,1,0,4.534456014633179
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,512,64,1,0,7.223828017711639
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,512,128,1,0,13.145488023757935
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1,1,1,0,0.7711599990725517
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1,4,1,0,0.7746679931879044
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,512,256,1,0,25.85732388496399
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1,8,1,0,0.70363999158144
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1,16,1,0,0.5993559882044792
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1,32,1,0,0.5945480018854141
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1,64,1,0,0.5849999934434891
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1024,4,1,0,4.432672023773193
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1,128,1,0,0.6010560095310211
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1,256,1,0,0.6100759878754616
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1,512,1,0,0.6556079983711243
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1,1024,1,0,0.7689839899539948
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1024,8,1,0,5.063040018081665
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1,2048,1,0,1.1958879977464676
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1,4096,1,0,2.3980640172958374
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1024,16,1,0,6.262475967407227
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1,8192,1,0,6.640572011470795
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1024,32,1,0,8.915748000144958
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1,16384,1,0,20.823835849761963
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1024,64,1,0,14.444387912750244
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,2,4,1,0,0.6041040048003197
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,2,8,1,0,0.6172079965472221
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,2,16,1,0,0.6048679947853088
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,2,32,1,0,0.6042479947209358
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,2,64,1,0,0.6046520099043846
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,2,128,1,0,0.6269640177488327
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,2,256,1,0,0.6635960042476654
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,2,512,1,0,0.7439119964838028
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,2,1024,1,0,1.1059480011463165
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,2,2048,1,0,1.9909000098705292
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,2,4096,1,0,4.806831955909729
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,64,1024,128,1,0,26.01899218559265
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,2,8192,1,0,13.618640065193176
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,2,16384,1,0,42.1333122253418
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,4,4,1,0,0.6122159957885742
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,4,8,1,0,0.6110479980707169
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,4,16,1,0,0.6101519986987114
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,4,32,1,0,0.9627599939703941
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,4,64,1,0,0.987292006611824
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,4,128,1,0,1.0214999914169312
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,4,256,1,0,1.1229360103607178
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,4,512,1,0,1.3160320073366165
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,4,1024,1,0,1.7834759801626205
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,4,2048,1,0,3.867780029773712
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,4,4096,1,0,9.826204180717468
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,4,8192,1,0,27.180628061294556
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,4,16384,1,0,87.511962890625
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,8,4,1,0,0.6062759906053543
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,8,8,1,0,0.6032519936561584
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,8,16,1,0,0.6153199896216393
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,8,32,1,0,0.6226240023970604
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,8,64,1,0,0.6626079976558685
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,8,128,1,0,0.7409639954566956
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,8,256,1,0,1.0257359892129898
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,8,512,1,0,1.6907079964876175
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,8,1024,1,0,3.385692000389099
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,8,2048,1,0,7.698932111263275
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,8,4096,1,0,19.42417597770691
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,8,8192,1,0,54.656972885131836
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,8,16384,1,0,175.3879108428955
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,16,4,1,0,0.6215399950742722
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,16,8,1,0,0.627255991101265
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,16,16,1,0,0.64376400411129
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,16,32,1,0,0.6778799965977669
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,16,64,1,0,0.7869920060038567
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,16,128,1,0,1.0518879890441895
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,16,256,1,0,1.7008279860019684
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,16,512,1,0,3.3014639616012573
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,16,1024,1,0,7.159072041511536
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,16,2048,1,0,15.773539900779724
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,16,4096,1,0,40.21097993850708
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,32,4,1,0,0.7434680014848709
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,32,8,1,0,0.7602960020303726
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,32,16,1,0,1.1946399807929993
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,32,32,1,0,0.8716040030121803
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,32,64,1,0,1.1329519897699356
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,32,128,1,0,1.6608919948339462
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,32,256,1,0,3.1008640229701996
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,32,512,1,0,6.39357602596283
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,32,1024,1,0,13.898675918579102
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,32,2048,1,0,31.692668199539185
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,32,4096,1,0,80.60142707824707
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,64,4,1,0,0.9376679956912994
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,64,8,1,0,0.9871640130877495
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,64,16,1,0,1.0389520227909088
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,64,32,1,0,1.2701719850301743
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,64,64,1,0,1.8133839964866638
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,64,128,1,0,3.124444007873535
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,64,256,1,0,6.156515955924988
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,64,512,1,0,12.935196042060852
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,64,1024,1,0,27.650256395339966
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,64,2048,1,0,63.94277238845825
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,128,4,1,0,1.3317159861326218
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,128,8,1,0,1.3820279836654663
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,128,16,1,0,1.6104319840669632
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,128,32,1,0,2.139351963996887
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,128,64,1,0,3.4658039808273315
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,128,128,1,0,6.2845799922943115
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,128,256,1,0,12.173516035079956
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,128,512,1,0,25.73851990699768
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,128,1024,1,0,58.33561563491821
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,256,4,1,0,2.112404018640518
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,256,8,1,0,2.302264004945755
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,256,16,1,0,2.823412001132965
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,256,32,1,0,4.105195999145508
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,256,64,1,0,6.766952037811279
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,256,128,1,0,12.441127896308899
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,256,256,1,0,24.780704021453857
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,256,512,1,0,54.597976207733154
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,512,4,1,0,3.6549719870090485
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,512,8,1,0,4.139860033988953
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,512,16,1,0,5.413680076599121
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,512,32,1,0,7.895468056201935
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,512,64,1,0,13.404811978340149
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,512,128,1,0,24.85066795349121
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,512,256,1,0,52.19697666168213
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1024,4,1,0,7.2684319615364075
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1024,8,1,0,8.097892045974731
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1024,16,1,0,10.571960091590881
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1024,32,1,0,16.221920013427734
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1024,64,1,0,27.011167764663696
SGLang,0.5.6.post2,NVIDIA H200,mla_context,fa3,fp8_block,fp8,128,1024,128,1,0,51.77046775817871
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1,1,1,0,0.6067040041089058
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1,1,1,0,0.6079559996724129
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1,4,1,0,0.5988399982452393
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1,1,1,0,0.6076639965176582
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1,4,1,0,0.6098680049180984
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1,1,1,0,0.6177880018949509
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1,8,1,0,0.5910560041666031
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1,4,1,0,0.590816006064415
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1,8,1,0,0.6033640056848526
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1,4,1,0,0.6157039999961853
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1,16,1,0,0.605492003262043
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1,8,1,0,0.5950079932808876
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1,16,1,0,0.6128560081124306
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1,8,1,0,0.6059840023517609
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1,32,1,0,0.6027839928865433
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1,16,1,0,0.5973600074648857
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1,32,1,0,0.5937799960374832
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1,16,1,0,0.7625919952988625
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1,64,1,0,0.5887119919061661
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1,32,1,0,0.5937000066041946
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1,64,1,0,0.595100000500679
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1,32,1,0,0.5958279967308044
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1,128,1,0,0.5895599946379662
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1,64,1,0,0.5826759934425354
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1,128,1,0,0.5825719982385635
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1,64,1,0,0.614876002073288
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1,256,1,0,0.5945480093359947
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1,128,1,0,0.5843600034713745
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1,256,1,0,0.6089920029044151
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1,128,1,0,0.5997559875249863
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1,512,1,0,0.59757199883461
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1,256,1,0,0.5973520055413246
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1,512,1,0,0.6461319997906685
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1,256,1,0,0.6048559993505478
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1,1024,1,0,0.611467994749546
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1,512,1,0,0.6298520117998123
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1,1024,1,0,0.7677880078554153
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1,512,1,0,0.6011639907956123
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1,2048,1,0,0.6999080181121826
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1,1024,1,0,0.6640599966049194
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1,1024,1,0,0.6208479925990105
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1,2048,1,0,1.1822439879179
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1,4096,1,0,0.928571991622448
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1,2048,1,0,0.8576159924268723
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1,2048,1,0,0.6317279934883118
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1,4096,1,0,0.7376840114593506
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1,4096,1,0,1.3563720136880875
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1,8192,1,0,1.818892002105713
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1,4096,1,0,2.4214119911193848
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1,8192,1,0,1.1823080033063889
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1,8192,1,0,3.3361200392246246
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1,16384,1,0,2.6893639862537384
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1,16384,1,0,4.964303910732269
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,2,4,1,0,0.604483999311924
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,2,4,1,0,0.6006999984383583
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1,8192,1,0,6.888779938220978
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,2,8,1,0,0.6169440075755119
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,2,8,1,0,0.6058440059423447
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,2,16,1,0,0.6021840050816536
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,2,16,1,0,0.6101559996604919
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,2,32,1,0,0.6055880039930344
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,2,32,1,0,0.6052400022745132
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,2,64,1,0,0.6206640005111694
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,2,64,1,0,0.5916360020637512
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,2,128,1,0,0.6023039892315865
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1,16384,1,0,10.682071924209595
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,2,128,1,0,0.6006160005927086
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,2,256,1,0,0.6077599972486496
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,2,4,1,0,0.5970680043101311
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,2,256,1,0,0.6061679944396019
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,2,512,1,0,0.6301440075039864
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,2,8,1,0,0.6097039952874184
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,2,512,1,0,0.645687997341156
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,2,1024,1,0,0.6553720086812973
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,2,16,1,0,0.5989000052213669
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,2,1024,1,0,0.6768360063433647
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,2,2048,1,0,0.6926079988479614
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,2,32,1,0,0.5939039960503578
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,2,2048,1,0,0.8305879980325699
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,2,4096,1,0,0.9626720026135445
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,2,64,1,0,0.597960002720356
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,2,4096,1,0,1.3831759840250015
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,2,128,1,0,0.6078399941325188
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,2,8192,1,0,1.8815719932317734
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,2,256,1,0,0.6233119890093803
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,2,512,1,0,0.6724800020456314
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,2,8192,1,0,3.350243955850601
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1,16384,1,0,21.483416080474854
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,2,1024,1,0,0.7827439904212952
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,2,4,1,0,0.6053480058908463
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,2,16384,1,0,5.282136023044586
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,2,2048,1,0,1.1272040009498596
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,2,8,1,0,0.5957559943199158
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,4,4,1,0,0.6073759943246841
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,2,16,1,0,0.5918200016021729
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,4,8,1,0,0.6131839901208878
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,2,4096,1,0,2.480811983346939
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,2,32,1,0,0.6028400138020515
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,4,16,1,0,0.6164400056004524
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,2,64,1,0,0.592616006731987
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,4,32,1,0,0.6264360025525093
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,2,128,1,0,0.630463995039463
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,2,16384,1,0,10.538652062416077
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,4,64,1,0,0.6159960031509399
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,2,256,1,0,0.6636280044913292
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,4,4,1,0,0.6164839938282967
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,4,128,1,0,0.6175919994711876
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,2,8192,1,0,6.775884032249451
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,2,512,1,0,0.7884120047092438
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,4,8,1,0,0.613307997584343
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,4,256,1,0,0.6173279955983162
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,2,1024,1,0,1.079012006521225
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,4,16,1,0,0.6017479971051216
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,4,512,1,0,0.6396439969539642
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,4,32,1,0,0.6164640039205551
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,4,1024,1,0,0.6674319952726364
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,2,2048,1,0,1.9965039938688278
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,4,2048,1,0,0.8482760116457939
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,4,64,1,0,0.609420008957386
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,4,128,1,0,0.6346840038895607
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,4,4096,1,0,1.4520799964666367
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,4,256,1,0,0.658500000834465
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,2,4096,1,0,4.8785359263420105
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,2,16384,1,0,21.457056045532227
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,4,512,1,0,0.7004679962992668
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,4,4,1,0,0.6037760078907013
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,4,1024,1,0,1.1621080040931702
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,4,8192,1,0,3.573607951402664
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,4,8,1,0,0.5980240032076836
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,4,16,1,0,0.6104200035333633
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,4,2048,1,0,1.1640959978103638
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,4,32,1,0,0.5995839908719063
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,4,64,1,0,0.6019959971308708
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,4,4096,1,0,2.5072519779205322
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,4,128,1,0,0.6409559994935989
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,4,256,1,0,0.6733400002121925
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,4,512,1,0,0.7622079998254776
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,2,8192,1,0,13.897815823554993
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,4,1024,1,0,1.024464026093483
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,4,16384,1,0,11.276856064796448
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,4,8192,1,0,6.614424049854279
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,8,4,1,0,0.6134759932756424
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,4,2048,1,0,2.0637039840221405
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,8,8,1,0,0.6170879900455475
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,8,16,1,0,0.619391992688179
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,8,32,1,0,0.6138719990849495
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,8,64,1,0,0.6105640009045601
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,4,4096,1,0,4.692972004413605
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,8,128,1,0,0.7099360153079033
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,8,256,1,0,0.6474080085754395
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,8,512,1,0,0.6794039979577065
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,8,1024,1,0,0.7951880022883415
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,8,2048,1,0,1.229663997888565
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,8,4096,1,0,2.7010479867458344
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,4,8192,1,0,13.95337188243866
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,4,16384,1,0,21.74315595626831
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,8,4,1,0,0.6125400066375732
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,8,8,1,0,0.6239480078220367
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,8,16,1,0,0.6154320016503334
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,8,8192,1,0,7.2296560406684875
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,8,32,1,0,0.6140479892492294
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,8,64,1,0,0.6252320036292076
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,8,128,1,0,0.6427239999175072
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,8,256,1,0,0.6860200092196465
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,8,512,1,0,0.7481679990887642
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,8,1024,1,0,1.0736520141363144
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,2,16384,1,0,42.717047691345215
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,8,2048,1,0,2.0916559994220734
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,4,4,1,0,0.6136199980974197
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,4,8,1,0,0.6036240011453629
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,4,16,1,0,0.6023079976439476
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,8,4096,1,0,4.86002391576767
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,4,32,1,0,0.6098880022764206
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,4,64,1,0,0.6191319972276688
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,4,128,1,0,0.6657639965415001
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,4,256,1,0,0.7455279976129532
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,8,16384,1,0,22.74780035018921
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,4,512,1,0,1.0309800058603287
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,16,4,1,0,0.6307799890637398
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,16,8,1,0,0.6395759955048561
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,4,1024,1,0,1.774152010679245
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,16,16,1,0,0.6264999955892563
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,16,32,1,0,0.635067991912365
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,16,64,1,0,0.6358160004019737
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,4,2048,1,0,3.811323970556259
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,8,8192,1,0,13.997895956039429
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,16,128,1,0,0.6606200039386749
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,16,256,1,0,0.6841719970107079
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,4,16384,1,0,43.58252763748169
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,16,512,1,0,0.7740360051393509
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,8,4,1,0,0.6123519986867905
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,16,1024,1,0,1.1265239864587784
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,8,8,1,0,0.6114799976348877
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,8,16,1,0,0.6059879958629608
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,16,2048,1,0,2.265032023191452
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,4,4096,1,0,9.584156036376953
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,8,32,1,0,0.6112399995326996
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,8,64,1,0,0.6331320032477379
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,8,128,1,0,0.6785319894552231
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,8,256,1,0,0.7637320011854172
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,16,4096,1,0,5.286244094371796
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,8,512,1,0,0.9805919975042343
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,32,4,1,0,0.6676320061087608
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,32,8,1,0,0.6674559935927391
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,8,1024,1,0,1.8476840108633041
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,32,16,1,0,0.6684159860014915
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,32,32,1,0,0.6634519919753075
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,32,64,1,0,0.7010280042886734
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,8,2048,1,0,3.9145920276641846
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,32,128,1,0,0.7238800004124641
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,32,256,1,0,0.8181559965014458
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,32,512,1,0,1.1176719963550568
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,32,1024,1,0,2.076740026473999
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,8,4096,1,0,10.022804021835327
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,4,8192,1,0,27.519960165023804
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,32,2048,1,0,4.414379954338074
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,8,16384,1,0,43.535096168518066
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,16,4,1,0,0.6323160007596016
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,16,8,1,0,0.6204759925603867
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,16,16,1,0,0.6276679933071136
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,16,32,1,0,0.6411359906196594
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,32,4096,1,0,10.821995973587036
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,16,64,1,0,0.6617079973220825
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,64,4,1,0,0.7331880033016205
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,16,128,1,0,0.7009199857711792
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,64,8,1,0,0.7164719998836517
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,16,256,1,0,0.7614999935030937
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,64,16,1,0,0.7329840064048767
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,16,512,1,0,1.0222679823637009
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,64,32,1,0,0.7404839992523193
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,64,64,1,0,0.7877440005540848
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,16,1024,1,0,1.8910760134458542
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,64,128,1,0,0.8642840012907982
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,8,8192,1,0,27.571571826934814
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,64,256,1,0,1.1357200145721436
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,16,2048,1,0,3.9879279732704163
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,64,512,1,0,2.0189360082149506
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,64,1024,1,0,3.9843920171260834
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,16,4096,1,0,10.178007960319519
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,32,4,1,0,0.6467960029840469
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,32,8,1,0,0.6570760011672974
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,32,16,1,0,0.6675360053777695
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,64,2048,1,0,9.063031911849976
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,32,32,1,0,0.6895599961280823
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,32,64,1,0,0.7241160050034523
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,128,4,1,0,0.8418920040130615
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,32,128,1,0,0.7934359982609749
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,128,8,1,0,0.8334039971232414
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,32,256,1,0,1.0460319966077805
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,128,16,1,0,0.8538400009274483
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,128,32,1,0,0.9155719950795174
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,32,512,1,0,1.8163039982318878
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,128,64,1,0,0.9747839942574501
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,128,128,1,0,1.2298279851675034
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,32,1024,1,0,3.57437601685524
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,128,256,1,0,2.075844019651413
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,128,512,1,0,3.85563200712204
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,32,2048,1,0,8.3020761013031
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,128,1024,1,0,8.239935994148254
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,256,4,1,0,1.0626680105924606
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,256,8,1,0,1.1031000018119812
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,256,16,1,0,1.1360480040311813
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,4,16384,1,0,87.16015625
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,256,32,1,0,1.2308600097894669
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,8,4,1,0,0.6088439971208572
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,8,8,1,0,0.6095039993524551
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,256,64,1,0,1.4660039842128754
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,8,16,1,0,0.6064120009541512
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,32,4096,1,0,20.11245608329773
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,8,32,1,0,0.625059999525547
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,256,128,1,0,2.2636919915676117
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,64,4,1,0,0.7104519903659821
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,8,64,1,0,0.6686320081353188
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,8,128,1,0,0.7565880045294762
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,64,8,1,0,0.7306080013513565
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,8,256,1,0,1.010503999888897
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,64,16,1,0,0.7439000010490417
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,256,256,1,0,3.9565120339393616
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,64,32,1,0,0.780236005783081
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,8,512,1,0,1.6742359846830368
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,64,64,1,0,0.844311997294426
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,64,128,1,0,1.0683719962835312
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,8,1024,1,0,3.386759966611862
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,64,256,1,0,1.838352009654045
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,256,512,1,0,7.793260037899017
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,64,512,1,0,3.4029840230941772
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,512,4,1,0,1.629620000720024
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,8,2048,1,0,7.812040030956268
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,8,16384,1,0,88.08416366577148
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,512,8,1,0,1.6547839939594269
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,16,4,1,0,0.6284399926662445
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,16,8,1,0,0.6182839870452881
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,512,16,1,0,1.7220679819583893
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,64,1024,1,0,7.137828052043915
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,16,16,1,0,0.6306040063500404
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,16,32,1,0,0.6490559950470924
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,512,32,1,0,1.94548399746418
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,16,64,1,0,0.689860001206398
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,16,128,1,0,0.7707319930195808
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,512,64,1,0,2.706415981054306
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,16,256,1,0,0.9714159891009331
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,16,512,1,0,1.7574960142374039
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,512,128,1,0,4.310496032238007
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,16,1024,1,0,3.419867992401123
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,8,4096,1,0,19.86317205429077
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,64,2048,1,0,16.50328016281128
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,128,4,1,0,0.8678719997406006
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,128,8,1,0,0.8741279914975166
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,16,2048,1,0,8.12387603521347
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,512,256,1,0,7.802851974964142
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,128,16,1,0,0.9511680006980896
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,128,32,1,0,0.997143991291523
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,128,64,1,0,1.21254400908947
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1024,4,1,0,2.7275959849357605
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,128,128,1,0,1.91457200050354
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1024,8,1,0,2.8309080004692078
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,128,256,1,0,3.412032037973404
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1024,16,1,0,3.0043319761753082
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,16,4096,1,0,20.038715600967407
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,128,512,1,0,6.783963978290558
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1024,32,1,0,3.787580043077469
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,32,4,1,0,0.6414599940180779
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,32,8,1,0,0.6629239991307259
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,32,16,1,0,0.6593599990010262
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,32,32,1,0,0.7251920029520988
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1024,64,1,0,5.309267938137054
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,32,64,1,0,0.8024400100111961
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,32,128,1,0,0.9836919978260994
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,32,256,1,0,1.7466880083084106
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,128,1024,1,0,14.738264083862305
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,32,512,1,0,3.2114839255809784
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,16,1024,128,1,0,8.521903991699219
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,256,4,1,0,1.2012480050325394
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,256,8,1,0,1.2459239959716797
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,256,16,1,0,1.3049200028181076
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,32,1024,1,0,6.7385119795799255
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,8,8192,1,0,57.93162393569946
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,256,32,1,0,1.5083640068769455
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,256,64,1,0,2.1868560314178467
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,256,128,1,0,3.563372015953064
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,32,2048,1,0,15.977855920791626
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,256,256,1,0,6.780175983905792
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,256,512,1,0,13.8959321975708
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,512,4,1,0,1.853620007634163
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,512,8,1,0,1.9319519996643066
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,512,16,1,0,2.099448025226593
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,512,32,1,0,2.7597320079803467
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,512,64,1,0,4.108443915843964
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,32,4096,1,0,40.634275913238525
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,64,4,1,0,0.7635999992489815
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,64,8,1,0,0.7732360064983368
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,512,128,1,0,7.091259956359863
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,64,16,1,0,0.8222959935665131
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,64,32,1,0,0.9290399923920631
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,64,64,1,0,1.080372005701065
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,64,128,1,0,1.7766840010881424
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,64,256,1,0,3.208027958869934
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,512,256,1,0,13.620568037033081
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,64,512,1,0,6.438696086406708
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1024,4,1,0,3.168091982603073
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1024,8,1,0,3.3662959933280945
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,64,1024,1,0,14.031720042228699
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1024,16,1,0,4.0691879987716675
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1024,32,1,0,5.3019280433654785
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1024,64,1,0,8.069808065891266
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,64,2048,1,0,32.391355991363525
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,128,4,1,0,1.0006240010261536
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,128,8,1,0,1.036227986216545
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,32,1024,128,1,0,14.272420048713684
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,128,16,1,0,1.1357119977474213
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,128,32,1,0,1.3000959903001785
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,128,64,1,0,1.9547000080347061
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,128,128,1,0,3.2558239698410034
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,128,256,1,0,6.358915984630585
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,128,512,1,0,13.309272170066833
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,8,16384,1,0,178.66693115234375
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,16,4,1,0,0.6199920028448105
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,16,8,1,0,0.6233120039105415
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,16,16,1,0,0.6394639909267426
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,16,32,1,0,0.6791200041770935
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,16,64,1,0,0.7744000032544136
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,16,128,1,0,1.0318719893693924
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,16,256,1,0,1.6549079716205597
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,16,512,1,0,3.1880880296230316
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,16,1024,1,0,6.95470803976059
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,128,1024,1,0,28.369376182556152
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,256,4,1,0,1.4595600068569183
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,256,8,1,0,1.5514119863510132
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,256,16,1,0,1.6801559776067734
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,256,32,1,0,2.3354039788246155
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,16,2048,1,0,15.729879975318909
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,256,64,1,0,3.5858120024204254
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,256,128,1,0,6.4314239621162415
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,256,256,1,0,12.788928151130676
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,16,4096,1,0,40.37076759338379
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,32,4,1,0,0.7265479937195778
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,32,8,1,0,0.7084719985723495
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,32,16,1,0,0.7616400048136711
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,32,32,1,0,0.852848008275032
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,32,64,1,0,1.0958719700574875
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,32,128,1,0,1.6647320091724396
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,256,512,1,0,26.54879593849182
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,32,256,1,0,3.120356023311615
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,512,4,1,0,2.3832879960536957
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,512,8,1,0,2.490359991788864
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,32,512,1,0,6.471831977367401
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,512,16,1,0,3.1479199826717377
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,512,32,1,0,4.376863896846771
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,32,1024,1,0,13.842943787574768
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,512,64,1,0,7.015483975410461
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,512,128,1,0,13.01911199092865
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,32,2048,1,0,31.701308012008667
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,512,256,1,0,25.999759674072266
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1024,4,1,0,4.201040089130402
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1024,8,1,0,4.8231600522994995
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1024,16,1,0,6.028271973133087
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1024,32,1,0,8.575064063072205
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1024,64,1,0,14.183032035827637
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,32,4096,1,0,83.53198051452637
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,64,4,1,0,0.8855599984526634
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,64,8,1,0,0.9112799987196922
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,64,1024,128,1,0,26.006139993667603
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,64,16,1,0,0.9991719946265221
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,64,32,1,0,1.2404840141534805
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,64,64,1,0,1.7966759949922562
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,64,128,1,0,3.149136006832123
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,64,256,1,0,6.205676019191742
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,64,512,1,0,13.037443995475769
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,64,1024,1,0,27.65142798423767
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,64,2048,1,0,65.4186840057373
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,128,4,1,0,1.3063520044088364
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,128,8,1,0,1.3316440135240555
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,128,16,1,0,1.5547119975090027
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,128,32,1,0,2.0800239741802216
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,128,64,1,0,3.379828006029129
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,128,128,1,0,6.2258999943733215
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,128,256,1,0,12.541859984397888
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,128,512,1,0,25.919935703277588
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,128,1024,1,0,58.32420825958252
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,256,4,1,0,2.0320920050144196
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,256,8,1,0,2.1673199832439423
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,256,16,1,0,2.6693080365657806
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,256,32,1,0,3.9588319659233093
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,256,64,1,0,6.645947992801666
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,256,128,1,0,12.467751860618591
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,256,256,1,0,25.124011993408203
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,256,512,1,0,54.455124378204346
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,512,4,1,0,3.4629800021648407
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,512,8,1,0,3.8496879935264587
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,512,16,1,0,5.128655970096588
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,512,32,1,0,7.7991639375686646
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,512,64,1,0,13.34091603755951
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,512,128,1,0,24.791388273239136
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,512,256,1,0,52.026512145996094
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1024,4,1,0,6.386503994464874
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1024,8,1,0,7.566923975944519
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1024,16,1,0,10.180675983428955
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1024,32,1,0,15.526984095573425
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1024,64,1,0,26.982152462005615
SGLang,0.5.6.post2,NVIDIA H200,mla_context,flashinfer,fp8_block,fp8,128,1024,128,1,0,51.25774383544922
